Bayesian Markov switching models for the early detection of influenza epidemics

Abstract

The early detection of the outbreaks of diseases is one of the most challenging objectives of epidemiological surveillance systems. In this paper, a Markov switching model is introduced to determine the epidemic and non-epidemic periods from influenza surveillance data: the process of differenced incidence rates is modelled either with a first-order autoregressive process or with a Gaussian white noise process depending on whether the system is in an epidemic or a nonepidemic phase. The transition between phases of the disease is modelled as a Markovian process. Bayesian inference is carried out on the former model to detect influenza epidemics at the very moment of their onset. Moreover, the proposal provides the probability of being in an epidemic state at any given moment. The methodology is evaluated on influenza illness data obtained from the Sanitary Sentinel Network of the Comunitat Valenciana, one of the 17 autonomous regions in Spain.

Discussion on “Optimal Sequential Surveillance for Finance, Public Health, and Other Areas” by Marianne Frisén

“…This is good for detecting a year that differs from other years, but this method will not detect an outbreak that comes later than usual. An approach for outbreak detection that does not involve estimation of the in-control level is suggested by Martínez-Beneito et al (2008), where the surveillance is based on the difference x t − x t − 1 and a detection of a shift from 0. To differentiate means that information is lost and that the time dependency structure changes (Petzold et al, 2004).…”

Improving regional influenza surveillance through a combination of automated outbreak detection methods: the 2015/16 season in France

“…We used a two-state hidden Markov model to differentiate between epidemic and non-epidemic weeks [15][16][17][18]. In our model, under state 1 (non-epidemic state), the observations were normally distributed with mean µ 1 and variance σ12; under state 2 (epidemic state), the observations were normally distributed with mean µ 2 and variance σ 22 .…”

Putting a Price on Climate Change: The Cost of Climate Change and Foodborne Diseases in Queensland in Quality-Adjusted Life Years and Dollars

“…Switching models simultaneously fit two models to a time-series and alternate between modelling sporadic and outbreak cases [124]. Outbreak cases are modelled using an AR-1 autoregressive term, reflecting the nature of outbreaks as inter-related, while sporadic cases are modelled using temperature and rainfall predictors.…”

“…Switching models are currently applied to infectious disease surveillance, such as in influenza monitoring in Spain [124]. The use of a switching model in this circumstance, to disentangle independent effects of weather variables from other seasonal influences, presents a novel application of this methodology which could be applied beyond the scope of foodborne diseases in future.…”

“…The model applied one of the two equations to predicting foodborne disease cases by determining shifts in the daily number of cases, with large changes incurring a phase change [124]. The probability of switching at time t depends on the state at time t − 1.…”